How cloud computing can improve 5G wireless networks
A great deal has been written about the technologies fueling 5G, especially how those technologies will improve the experience that users have regarding connectivity. Similarly, much has been said about how ongoing developments in technology will usher in a new generation of network-aware applications. In this article, we discuss one key aspect of 5G technology and how it will impact the development of wireless network capacity.
This is one of the more important but often neglected aspects of wireless communication evolution. It represents yet another important reason why the convergence of cloud computing and wireless communications makes so much sense. To state it briefly, many of the complex problems associated with 5G wireless networks can be solved using software, which sets aside the need for costly, time-intensive, and often slow-to-evolve hardware that’s been used in the past.
Cloud and telecommunication: A perfect fit for next-generation networks
It is well understood that some of the most sophisticated technologies that make up 5G can be implemented in software running on off-the-shelf servers. This is exciting because we can slowly but surely walk away from specialized hardware, which has been used in all the previous four generations of telecommunication networks. Transitioning to software will help reduce the overall capital and operational expenses for telecommunication providers. Equally important, this shift from hardware to software will make such networks future-proof because they will empower the telecommunications industry to become nimble and aggressive when rolling out desirable features at a regular cadence rather than having to wait a decade or so for the next generation standards to emerge. Innovation will thrive as we create a world where going from one generation to another is a software upgrade, just as the cloud industry has been doing for over a decade.
We will say more about this in future blogs, but today, we want to discuss wireless capacity—or technically, spectrum efficiency. Hopefully, we will convince you that computing power can be used to increase cellular network capacity, and advances in software-based machine learning and data analytics techniques can be used to improve the efficiency of 5G and future networks. When adding this to the other elements of the ecosystem the marriage between cloud computing and telecommunications networks makes a perfect fit.
5G core technologies: Massive multi-user MIMO
Multiple-input and multiple-output (MIMO) is a method for multiplying the capacity of a radio link, using multiple transmission, and receiving antennas to exploit multipath propagation. MIMO is an essential element of wireless communication standards in Wi-Fi, 3G, and 4G. 5G, however, takes it to the next level with massive multi-user (MU) MIMO, scaling the number of antennas massively and supporting many users simultaneously. This technology is the key to 5G’s promise of 1,000 times the capacity gain over 4G.
The science behind massive MU-MIMO lies in the complex mathematics involved with manipulating signals sent to and received from every antenna so that communication channels with each user can be preserved and can survive the environmental distortion. This has been the subject of many technical books and academic studies, but you can find a simplified version in the illustration below.
Massive MU-MIMO involves lots of matrix multiplications and transpositions, all of which require significant computation. It is a direct function of the number of users being serviced by the cell tower, and the number of antennas the cell tower has. Furthermore, this computation takes place every few milliseconds for thousands of subcarriers. The implication here is that significant processing power and energy are needed. As network operators increase the number of antennas, the computational requirement goes up considerably, along with other associated problems.
User patterns also affect the amount of computation needed. The precoding method described in the above figure works best if the users are stationary or moving slowly. Otherwise, the precoding matrix must be recomputed frequently, needing even more computations. An alternative method, known as “conjugate beamforming,” may work better in this case, but the number of antennas must far exceed the number of users and the wireless capacity is generally reduced.
So, the overall capacity that the network delivers is a direct function of how much computation power the operator is willing to purchase and deploy at each of its thousands of cell towers. Edge computing, which allows the ability to scale up computing easily, is perfect for this. Even if some operators don’t need lots of capacity immediately, it is still vital to understand if the network is to be built in such a way that it can be easily scaled up as the demand for network capacity grows.
Microsoft has invested heavily in computation technology that can deliver massive MU-MIMO for 5G networks. As early as 2012, Microsoft Research invested in a practical solution to implement MU-MIMO, using distributed pipelines with a rack of commodity servers (an edge data center) to meet timing specs and to scale to hundreds of antennas (the technique was state-of-the-art and a report was published at SIGCOMM 2013).
Deep learning for wireless capacity
5G is moving towards an open architecture, with many ways to optimize a network. While this approach increases complexity, deep learning techniques can be used to take on these complexities, which are typically beyond human abilities to solve. In the above case about precoding for massive MIMO, we can apply deep learning techniques to select an algorithm that would reduce energy consumption while minimizing reduction in capacity. Through predictive analytics and modern software that adapts to dynamic network loads, 5G networks can become smarter.
Microsoft has invested heavily in machine learning and AI and supported the work of world-leading experts in this area. And we are working on augmenting telecommunication networks by designing deep learning algorithms that include domain knowledge. In addition to the example above, we are actively investigating how deep learning techniques may be used to control transmission power to reduce interference, and thus increase capacity.
Continuous machine learning (powered by flexible edge computing to model the dynamic radio frequency environment and user mobility patterns), along with managing the signal processing pipeline, creates a tremendous value proposition for the telecommunications industry. This massive step forward empowers the rapid incorporation of research breakthroughs into the system—not only for the purpose of increasing wireless capacity, but also to improve the total operational efficiency of 5G networks.
Azure: Where edge computing, the cloud, and telecommunication operators come together
For more than 10 years, Microsoft has invested heavily in edge computing and is continuing to do so. In particular, Azure is working to provide computation close to the cell towers where it will benefit network operators the most, as they look to cost-effectively scale their network. Additionally, through its Azure for Operators initiative, Microsoft is continually working to enable new first and third-party solutions that further enhance and simplify edge computing, from network connectivity to on-demand compute, to complete orchestration.
Given Microsoft’s ability to scale computation up as much and as often as operators demand, the power of technology at the edge—including massive MU-MIMO—is the answer that telecommunication operators have been looking for. Azure is here to support telecommunication operators in meeting their goals for increasing capacity as the network grows and evolves. While telecommunication providers increase the number of antennas and cell towers, Microsoft’s ability to spin up servers at scale and to manage them from anywhere in the world makes Azure the perfect fit for 5G and beyond for telecommunication networks.
Read the Azure for Operators eBook to learn more.
Source: Azure Blog Feed